This webpage contains a selection of project highlights I have created in my recent professional career. Each section contains an overview of the project and its impact while subsequent tabs contain more detailed information and snapshots. Hover and click on the tabs to see the full scope of each project:
Data-Driven Goal Setting: provided data-driven goal setting support to Valor’s school leaders. This support included creating visualizations of historical data and visualizations of future projections for the purpose of goal-setting.
Automated Data Analysis: created a tool that automates data analysis for teachers, saving each teacher at least one hour of work per quarter (one hour, ~40 teachers, three quarters of analysis: ~120 hours of crucial teacher time saved).
Building Operations Dashboard: created a dashboard that allows school leaders to monitor building operations in real time.
Sample Project Plans: designed project plans for compliance with two Tennessee state laws.
Description: As part of my role, I provide data-driven goal setting support to Valor’s school leaders. This support includes creating visualizations of historical data and visualizations of future projections for the purpose of goal-setting.
Research & Analysis: To create the following goal setting visualizations, I drew on historical state testing data and used the R programming language to create the visualizations.
Strategic Planning: A particular challenge of this project was the absence of data during Covid-19 school closures and a dip in student test scores following the closures. We set a performance range for teacher goals to be from 3% growth from the previous year (low) to pre-pandemic performance (high). We also wanted to understand longitudinal student performance in math and how that changed over time (second visualization). This involved some data transformation so we could visualize the data in a way that was meaningful to school leaders.
Implementation & Results: School leaders shared the visualizations directly with teachers to guide goal-setting conversations for the school year.
The following visualizations were used to set goals for teachers for the 23-24 school year and to understand historical student performance.
R code for teacher goal setting
generate_plot <-function(plot_title, value_2021, value_2122, value_2223, flat_growth_val, goal_val, pre_pandemic_val, cohort_value_2021, cohort_value_2122, cohort_value_2223) {# Add the new line for cohort values df_cohort <-data.frame(School_Year =c("20-21", "21-22", "22-23"),Line_Type =rep("2) Cohort", 3),ELA_OTM =c(cohort_value_2021, cohort_value_2122, cohort_value_2223) ) df <-data.frame(School_Year =rep(c("20-21", "21-22", "22-23", "23-24"), each =4),Line_Type =rep(c("1) Grade Level", "3) Flat Growth", "4) Goal", "5) Pre-pandemic"), times =4), # Changed flat growth goal to just flat growthELA_OTM =c(value_2021, NA, NA, NA, value_2122, NA, NA, NA, value_2223, cohort_value_2223, cohort_value_2223, value_2223,NA, flat_growth_val, goal_val, pre_pandemic_val) )# Combine the new data with the existing dataframe df <-rbind(df, df_cohort) y_min <-25 y_max <-100# Update the plot p <-ggplot(df, aes(x = School_Year, y = ELA_OTM, group = Line_Type, color = Line_Type)) +geom_line(aes(linetype = Line_Type)) +geom_point(aes(color = Line_Type)) +geom_text(aes(label =ifelse(is.na(ELA_OTM), "", ELA_OTM), color = Line_Type), vjust =-1) +#geom_point(aes(color = ifelse(ELA_OTM %in% c(44.6, 46, 60.8), "Actual", Line_Type))) +geom_text(aes(label =ifelse(is.na(ELA_OTM), "", ELA_OTM), color =ifelse(ELA_OTM %in%c(value_2021, value_2122, value_2223), "1) Grade Level", Line_Type)), vjust =-1) +scale_y_continuous(limits =c(y_min, y_max), name ="% On Track and Mastered") +scale_x_discrete(name ="School Year") +######scale_linetype_manual(values = c("1) Grade Level" = "solid", "3) Flat Growth" = "dotted", "4) Goal" = "dotted", "5) Pre-pandemic" = "dotted", "2) Cohort" = "solid")) + # Changed flat growth goal to just flat growthguides(color =guide_legend(override.aes =list(shape =15, size=6))) +ggtitle(plot_title) +theme_minimal() +theme(text =element_text(family ="Times New Roman", size =12), legend.title =element_blank(), plot.title =element_text(hjust =0.5))# Save and printprint(p) filename <-paste0("../data/Valor_2324_Goals", plot_title)ggsave(filename, plot = p, width =6, height =4, dpi =300)}df_csv <-read.csv("../data/22-23 TNReady_EOC Scores - Graphing 23.24 Goals.csv")# Remove "%" from all cells in the DataFramedf_csv[] <-lapply(df_csv, function(x) gsub("%", "", x))# Coerce all columns except the first one to numericdf_csv[, -1] <-lapply(df_csv[, -1], function(x) as.numeric(as.character(x)))# Loop through each row of the dataframefor(i in1:nrow(df_csv)) {# Extract variables from current row plot_title <- df_csv[i, 1] value_2021 <- df_csv[i, 2] value_2122 <- df_csv[i, 3] value_2223 <- df_csv[i, 5] flat_growth_val <- df_csv[i, 6] goal_val <- df_csv[i, 7] pre_pandemic_val <- df_csv[i, 8] cohort_value_2021 <- df_csv[i, 9] cohort_value_2122 <- df_csv[i, 10] cohort_value_2223 <- df_csv[i, 11]# Generate plot using extracted variablesgenerate_plot(plot_title, value_2021, value_2122, value_2223, flat_growth_val, goal_val, pre_pandemic_val, cohort_value_2021, cohort_value_2122, cohort_value_2223)}
Individual course goal setting Writing one set of code (see above) allowed us to reproduce visualizations and create meaningful, individualized goal ranges for over 50 teachers. These goals took into account historical grade-level performance and the performance of the incoming cohort.
Historical network-level goal setting A second set of visualizations allowed us to set goals for the network as a whole and to evaluate historical performance. The visualization below shows the percentage of students at various state-designated levels of performance over time and how they transition between performance levels from grade to grade. We wanted to be able to answer questions such as–if a student falls behind, how likely are they to catch up? When during a student’s time at our school is most critical at predicting their eventual performance?
Description: Teachers previously filled out a version of the below information manually by clicking through many reports in our third-party assessment platform (the “see successes” tab and an abbreviated skill analysis). My personal goal was to virtually completely automate this process by only requiring teachers to download one report.
Research & Analysis: I relied on my personal experience of completing data analysis as a teacher to create an initial plan. I used the existing data analysis protocol as a starting point for the Google Sheets workbook. Finally, I knew that my data science skills had the potential to bring more efficiency to this quarterly process.
Strategic Planning: I knew that if I was going to ask the whole organization to make an adjustment, it would be important to make the process clear, logical, and easier than it was before. We worked out many bugs and obstacles by testing the tool with a subset of staff beforehand. We didn’t catch 100% of bugs, but I was able to assist teachers in the moment due to the strong testing and preparation we had completed beforehand.
Implementation & Results: With the new data analysis protocol, teachers can see patterns in student mastery along more dimensions. The tool saves each teacher at least one hour of work per quarter (one hour, ~40 teachers, three quarters of analysis: ~120 hours of crucial teacher time saved).
The setup page is the only page that requires action from teachers. I aimed to make it highly user-friendly and clear that they only needed to pull one type of report once.
The “see successes” tab has not undergone any changes from the previous data analysis protocol except that it is now automatic! This summary view was the first step in making meaning of high-level student mastery levels across the teacher’s class periods.
The skill analysis tab pulls any standards associated with the test questions from the teacher’s report and calculates the overall class % correct. Additionally, the sheet shows differentials in mastery for English learners and special education students. From here, teachers could fill in personal takeaways on the skill-level data—a level of granularity that previously required pulling many successive reports.
This view shows a concept-by-concept list of students by level of mastery. This list can be used by teachers to pull small groups for re-teaching of concepts. This view was not available previously and could hypothetically have been created through extensive manual data analysis.
Description: Valor’s chief operating officer wanted a simpler way to assess the condition of all four school buildings. It is important for the safe operation of the schools to be able to see which areas need intervention and what patterns exist. Additionally, seeing all data clearly can help with prioritizing which issues require immediate addressing and which issues can be responded to in a batched manner.
Research & Analysis: I began designing the dashboard by asking staff members on the operations team which aspects were most important to them. The team also provided me with a detailed rubric they occasionally used. I simplified the rubric into yes/no questions. Through several rounds of review, we were able to come to a final version of the operations walkthrough tool.
Strategic Planning: The first iterations of the project were in Google Data Studio. The team found that the tool was too slow to load and that it was difficult to make the tool user-friendly. To improve from this feedback, I transferred the tool to Google Sheets where the performance was much better.
Implementation & Results: As it stands currently, the team feels positively about the tool and checks it regularly. In the future, we will likely add a time series view to show operations performance over time.
The data source for the dashboards is this Google form. Scores for various sections of the Google form are binned into categories such as whether they pertain to orderliness/cleanliness/compliance and building area for the dashboard. Directors of operations at each building fill the entirety of the Google form out at least once per month. Below are a couple snapshots of the form:
The data studio view of the building was the first iteration of the Operations Dashboard. The team found this view helpful, but it was often slow to load filtered data. Still, the initial operations dashboard provided a clear building and location-based view of cleanliness, orderliness, and compliance that didn’t exist previously.
In response to the team’s feedback that they wanted a quicker-loading tool, I moved the database to Google Sheets and created a view of the last 30 days, a year-to-date view, a photographs view (where the team could view and share photos of unsatisfactory areas of the building), and a page for open-ended responses.
This page is identical to the view of the last 30 days, but provides a year-to-date view of scores around the building.
Staff members have the opportunity to upload photographs when indicators are not met which can be viewed on this tab by clicking or hovering on each of the links.
Finally, staff provide open-ended responses when indicators are not met which are viewable here.
Description: School leaders regularly rely on me for project planning guidance, accountability, and for developing metrics of success. A key ingredient of successful project planning at the organizational level is a backwards-planned list of actions and owners for each of those actions.
Research & Analysis: I was responsible for ensuring Valor’s compliance with two new Tennessee state laws focused on 1) computer science education and 2) expanding services for students who may have dyslexia. To understand our schools’ necessary actions, I began by reading the laws directly and interpreting them as best I could. I then read secondary resources published by the state and began to think about who on our team needed to be involved. Finally, as I am no lawyer, I ensured that we consulted legal counsel as part of the project plan.
Strategic Planning: Adding a full computer science course and screening many students for dyslexia would require involvement from many senior leaders and staff at various points. I “began with the end in mind” by asking myself what success would look like at the end of the project plan. From this vision, I could backwards plan to ensure we would achieve this vision. I included links to all of the relevant state resources, identified “need to know” people, and drafted the project plan. I shared the project plan with my manager and modified it according to her feedback.
Implementation & Results: The project plans have been in progress for several months and have, with a few exceptions and adjustments, remained on schedule.
Valor will offer a full-year computer science course to the all incoming ninth graders and will offer a quarter-long introductory course to all eighth graders.
Valor will screen all qualifying individuals in fifth through eighth grade for dyslexia and write individual learning plans for students meeting criteria for characteristics of dyslexia.
Description: I was selected to be the general manager of Georgetown University Data Science and Analytics’ summer bootcamp, an onboarding program for over 125 incoming graduate students. I had completed the bootcamp the summer prior and was excited to create a welcoming and strong preparatory experience for the masters program.
Research & Analysis: to begin, the two lead teaching assistants and I created a Google form survey to send to the full TA team about strengths and areas for growth of the bootcamp program. The most striking piece of feedback we received was that, while the bootcamp was helpful, there weren’t consistent grading practices and TAs had not been responsive in a timely matter to student questions. We also learned from the graduate program director that an important priority should be to improve student retention.
Strategic Planning: Building on our team’s and our program director’s feedback, I set a vision to create clear guidelines for TA responsiveness and a strong system for grading. For TA responsiveness, I drafted a set of guidelines and collected feedback from the team until we reached an agreed-upon version. We also decided to expand office hours by each committing one weekly hour to be available to students on zoom. Finally, I thought that creating a checklist grading spreadsheet that generated a feedback email would dually address students’ concerns about grading inconsistency and create an efficiency for the TA team.
Implementation & Results: We received strong feedback from the incoming class about their bootcamp experience through a post-completion survey. TAs also shared that they appreciated the checkbox rubric. This allowed them to clearly see if certain students were falling behind the pack and showed which concepts in the assignments could be reinforced in office hours. We achieved about the same level of student retention as previous years and believe there are other possibilities for boosting this metric (e.g., examining the admissions process, other community building activities through the bootcamp, etc.).
The image below shows one week of the TA grading spreadsheet. The spreadsheet contains checkboxes for each sub-component of the assignment. Students needed 80% or more of the checkboxes to “pass” for a week, but they could revise as many times as necessary. The last column contains automatically generated emails for students with their performance on the assignment and information about whether—and which components exactly—they needed to revise. I created the customized emails using Google’s JavaScript console.
The TA team created a “student status” page that pulled from all of the grading weeks to generate a summary view of student progress.
We also created a pivot table to show a summary view of student progress throughout the bootcamp. This pivot table will inform further revisions to the bootcamp curriculum and structure. We can now answer questions with quantitative evidence (i.e., when do students begin to drop off? Which weeks had the lowest pass rates?, etc.).
Description: As the executive assistant to the principal at Valor College Prep, I developed a staff duty schedule to ensure student safety, aiming to distribute 127 weekly duty hours among 85 staff while addressing their time preferences and operational needs. I also created a staff-facing view of the schedule so staff could see their own schedule and the schedule of their colleagues.
Research & Analysis: I surveyed staff for time preferences and consulted with culture, operations, and academic departments for schedule conflicts, leading to insights about balancing duty distribution with staff availability and school needs.
Strategic Planning: Referencing all of the data I collected, I devised a strategy to align staff duties with their preferences and school safety needs, creating a detailed matrix to manage conflicts and preferences, demonstrating adaptability in addressing logistical challenges.
Implementation & Results: I successfully implemented the duty schedule with stakeholder feedback, achieving 74% of staff preferences and maintaining operational efficiency, underscoring the importance of early planning and stakeholder consultation in logistical success.
The duties sheet served as the back-end for coordinating all schedules for staff. Staff who had school-based conflicts would have sections of the schedule blacked out so I wouldn’t schedule over them. Total minutes were calculated along with total staff members needed per post. I wanted to satisfy as many staff members’ first preferences I could. Staff filled out a survey of their preferences for times before I made the schedule, and the preferences were denoted by color codes. By the end, I was able to get to 74%, which I also tracked in this sheet.
Staff have access to all views of the duties workbook—from the underlying assumptions to the full sheet (previous tab), however, this tab is designed to be staff-facing. Staff can see their own schedule and the schedule of their colleagues.
